Why Use ggplot?

Might be some disadvantages:

Functions and Arguments

Let’s build some graphs

Now we need to load some packages The ggplot2 package is required to create graphs.

library(ggplot2)

Now, we need some data. Data are from NCAA history: NCAA_Champions

https://www.ncaa.com/history/basketball-men/d1
https://www.ncaa.com/history/basketball-women/d1

GGplot likes data organized as data frames with factors for groupings. We will add a column for gender then combine the data frames together.

mTable <- read.csv("../data/Bballmens.csv",header=TRUE)
mTable$gender = "Men"
wTable <- read.csv("../data/BballWomens.csv",header=TRUE)
wTable$gender = "Women"

Basketball <- rbind(mTable, wTable)
colnames(Basketball) <- c("Year","Champion","Coach","Score","RunnerUp","Site","Gender")

Let’s convert some of those data types into useful things:

Basketball$Champion <- as.character(as.factor(Basketball$Champion))
Basketball$Champion <- substr(Basketball$Champion, 1, nchar(Basketball$Champion) - 7)

Basketball$WinningScore <- substr(Basketball$Score, 1, 2)
Basketball$LosingScore <- substr(Basketball$Score,4,5)

Basketball$WinningScore<-as.numeric(Basketball$WinningScore)
Basketball$LosingScore<-as.numeric(Basketball$LosingScore)
Basketball$Year<- as.numeric(Basketball$Year)

Basketball <- subset(Basketball, Champion != "UNLV")  #Note: I removed UNLV because they scored over #100 points and I didn't want to reformat.

Basketball$Site <- sub(",.*$", "", Basketball$Site)

head(Basketball)
##   Year       Champion           Coach      Score       RunnerUp
## 1 2019       Virginia    Tony Bennett 85-77 (OT)     Texas Tech
## 2 2018      Villanova      Jay Wright      79-62       Michigan
## 3 2017 North Carolina    Roy Williams      71-65        Gonzaga
## 4 2016      Villanova      Jay Wright      77-74 North Carolina
## 5 2015           Duke Mike Krzyzewski      68-63      Wisconsin
## 6 2014    Connecticut     Kevin Ollie      60-54       Kentucky
##           Site Gender WinningScore LosingScore
## 1  Minneapolis    Men           85          77
## 2  San Antonio    Men           79          62
## 3      Phoenix    Men           71          65
## 4      Houston    Men           77          74
## 5 Indianapolis    Men           68          63
## 6    Arlington    Men           60          54

geoms and aesthetics

The basic component of ggplot graphs are geoms. Each geom requires at minimum a data frame from which to draw data, and specification of aesthetics. Aesthetics, aes(), require at minimum an x and y series to plot. Within aes(), you can also group by category, color by category, fill by category, and calculate basic stats by group. aes() is extremely important as it allows you to customize your plot by mapping graphical attributes to data. Anything specified in aes() is joined to your data and appears in legends.

All plots start with ggplot()

Here is a scatterplot of winning scores vs. losing scores from the Basketball dataset. Note: all plots start with ggplot(). Use + to iteratively add new features.

Scatter <- ggplot() +  
  geom_point(data = Basketball, aes(x=WinningScore, y = LosingScore))
Scatter

Add new features using ‘+’

Change color of points:

Scatter2 <- ggplot() +  
  geom_point(data = Basketball, aes(x = WinningScore, y = LosingScore), color="purple") 
Scatter2

Notice that color is specified outside of aes().This tells ggplot to apply “purple” to all points. When assigning colors outside aes(), you will use "“. Do not use”" for anything within aes() that you want connected to data, this will disconnect from the data.

Color by Factor

But, if don’t want all point purple and want to color by whether it was a Men’s or Women’s game, we specify color WITHIN aes().:

ScatterGroup <- ggplot() +  
  geom_point(data = Basketball, aes(x = WinningScore, y = LosingScore, color = Gender)) 
ScatterGroup

geom_line:

TimeSeries <- ggplot()+
  geom_line(data = Basketball, aes(x = Year, y = WinningScore, color = Gender))
TimeSeries

geom bar

If we only want to graph a subset of the data frame, in this case, frequent championship winner, we can use the base R subset() function within ggplot.

Bar <- ggplot(subset(Basketball, Champion %in% c("North Carolina","Duke","Connecticut","Tennessee","UCLA","Kentucky","Indiana","Lousiville")))+
  geom_bar(aes(x = Champion)) 
Bar

Bar plot of winning scores:

Bar2<-ggplot()+
  geom_bar(data = Basketball, aes(x = WinningScore)) 
Bar2

geom_histogram with stat=“count” is identical to geom_bar:

Hist <- ggplot() +
  geom_histogram(data = Basketball, aes(x = WinningScore), stat="count")  
## Warning: Ignoring unknown parameters: binwidth, bins, pad
Hist

Hist2 <- ggplot()+
  geom_histogram(data = Basketball, aes(x = WinningScore))
Hist2
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

We can change binwidth:

Hist3 <- ggplot() +
  geom_histogram(data=Basketball, aes(x=WinningScore),binwidth=10) 

Hist3

Group by category

Hist4 <- ggplot() +
  geom_histogram(data = Basketball, aes(x = WinningScore, fill = Gender), binwidth=3) 

Hist4

Above we use fill to group. For histogram or any areal geoms, fill specifies inside color, color specifies the border. Example:

Hist5 <- ggplot() +
  geom_histogram(data = Basketball, aes(x = WinningScore, fill = Gender), color = "black") 
Hist5
## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

Density Plots

DensPlot <- ggplot() +
  geom_density(data = Basketball, aes(x = WinningScore), fill = "green", alpha = 0.3) + #alpha is a transparency command (0 totally transparent to 1 totally opaque)
  geom_density(data = Basketball, aes(x = LosingScore), fill = "red", alpha = 0.3)
DensPlot

geom_violin and geom_boxplot

Boxplot <- ggplot(subset(Basketball, Champion %in% c("North Carolina","Duke","Connecticut","Tennessee","UCLA","Kentucky","Indiana","Lousiville"))) +
  geom_boxplot(aes(x = Champion, y = WinningScore))
Boxplot

OR

Violin <- ggplot() +
  geom_violin(data=Basketball, aes(x=Gender, y=WinningScore), fill="red", alpha=0.4) #alpha is a transparency setting
Violin

Combine Plot Types

Violin + boxplot

Violin + geom_boxplot(data=Basketball, aes(Gender, WinningScore), fill=NA, width=.5)

Scales

#You can manually adjust scales:

Scatter+
  scale_x_continuous(name="Winning Score", limits=c(30,100), breaks=c(30,40,50,60,70,80,90,100)) +
  scale_y_continuous(name="Losing Score", limits=c(30,100),breaks=c(30,40,50,60,70,80,90,100)) +
  labs(title="NCAA Championships Winning vs Losing Scores")

#### Labels We can change labels to anything you might want. Here, using the to add line break to axis label. Using scale_discrete for team names and scale_y_continous for scores

Boxplot2 <- Boxplot +
  scale_x_discrete(name="NCAA basketball champion",labels=c("Connecticut\nHuskies","Duke\nBlue Devils","Indiana\nHoosiers","Kentucky\nWildcats","Louisville\nCardinals","Maryland\nTerrapins","North Carolina\nTarHeels","Tennessee\nVolunteers","UCLA\nBruins")) + 
  scale_y_continuous(name="Winning Scores") 
Boxplot2

Colors And Legends

Before, when plots had groups on them legends were added automatically. For example, see the TimeSeries graph grouped by gender. ggplot added a default legend with default colors. You can change these colors to something different:

BetterColors <- c("brown4", "royalblue4")
TimeSeries + scale_color_manual(values=BetterColors)

Colors are controlled in scale_color_manual. For fills, use scale_fill_manual.

But, the above only works for “grouped” data. For example, when we made the Density plot that had both Winning Scores and Losing Scores (DensPlot), these are not grouped as each is a separate variable. If we want to control the colors of both density curves and get those changes to appear in a legend, we have to do it manually by mapping to aes(). See below:

DensPlot2 <- ggplot() +
  geom_density(data=Basketball, aes(x=WinningScore, fill="WinningPoints"), alpha=0.3) + 
  geom_density(data=Basketball,aes(x=LosingScore,fill="LosingPoints"), alpha=0.3) +
  scale_fill_manual(name="Legend", values = c(WinningPoints = "brown", LosingPoints = "midnightblue"),labels=c("Losers","Winners")) + 
  xlab("Points") +
  ylab("Density")

DensPlot2

“WinningPoints” is in aes(). This will map it to a scale. But, it is in parentheses to indicate not a variable. Here, the two fills we put in aes() earlier are mapped to scale_fill and we can make a name, specify colors, and labels for the legend.

Put mean line in density plots

DensPlot2 +
  geom_vline(data=Basketball, aes(xintercept=mean(WinningScore), na.rm=T), color="brown", size=0.8, linetype = "dashed") +
   geom_vline(data=Basketball, aes(xintercept = mean(LosingScore), na.rm=T), color="midnightblue", size=0.8, linetype="dashed")
## Warning: Ignoring unknown aesthetics: na.rm

## Warning: Ignoring unknown aesthetics: na.rm

Scale_color_gradient

ggplot() +  
  geom_point(data=Basketball,aes(x=WinningScore,y=LosingScore,color=Year),size=10) +
  scale_color_gradient(name="Year", low="black", high="red")

Some basic statistical visualizations (huge advantage of gg)

Basic arithmetic transformations:

Line2 <- ggplot() +
  geom_line(data=Basketball,aes(x=Year, y=WinningScore-LosingScore, color=Gender)) +
  scale_color_manual(values=BetterColors) +
  ylab("Point Differential")

Line2

Smooth the time series graph to make it easier to read

Line3 <- ggplot() +
  geom_smooth(data=Basketball, aes(x=Year, y=WinningScore-LosingScore, color=Gender), se=F) 

Line3
## `geom_smooth()` using method = 'loess' and formula 'y ~ x'

What does the se=F argument mean?

Lines of best fits:

Scatter2 +
  geom_smooth(data=Basketball, aes(WinningScore, LosingScore), method="lm")

Get a separate linear fit by group:

WinningScore <- ScatterGroup + geom_smooth(data=Basketball, aes(WinningScore, LosingScore,color=Gender), method="lm", se=FALSE, size=1) 
WinningScore

Themes

You can save themes for later use.

Theme1 <- theme(panel.border=element_rect(color="black",fill=NA,),panel.grid.minor=element_blank())
Theme2 <- theme(panel.border=element_rect(color="black",fill=NA),panel.background=element_blank())
Theme3 <- theme(panel.border=element_rect(color="black",fill=NA),panel.background=element_blank(), axis.text.x=element_text(size=11,angle=90,color="midnightblue"))

Boxplot2 + Theme1

Boxplot2 + Theme2

Boxplot2 + Theme3

Facets

Sites <- subset(Basketball, Site=="Houston"|Site=="Indianapolis"|Site=="Atlanta"|Site=="San Antonio"|Site=="New Orleans"|Site=="St. Louis")

Facet1 <- ggplot() +
  geom_density(data=Sites, aes(x=WinningScore), fill="black", alpha=0.4) +
  facet_wrap(~Site,ncol=2) + Theme2 +
  labs(title="Winning Scores by Site",x="Winning Score")

Facet1

Since it won’t always make sense for all facets to have identical scales, you can use scales=“free” to give each facet an appropriate scale. You can also use individualized themes within the ggplot call

Facet2 <- ggplot() +
  geom_density(data=Sites, aes(x=WinningScore), fill="black", alpha=0.2, linetype="dashed") +
  facet_wrap(~Site,ncol=2,scales="free_y") +
  Theme2+theme(strip.background=element_blank()) +
  labs(title="Winning Scores by Site",x="Winning Score")

Facet2

Saving plots

ggsave("WinningScore", plot = WinningScore, device = "jpeg", path = ".",
  scale = 1, width = 7, height = 5, units = "cm",
  dpi = 800, limitsize = TRUE)

Further Information

There is plenty of information on the web regarding ggplot2 with so many examples and community input. For example: Cran Manual

Some basics with geoms and aesthetics: GEOM_POINT geom_point help page

GEOM LINE geom_line

GEOM BOXPLOT geom_boxplot

VIOLIN PLOT geom_violin

SCALES continuous scales

COLORS AND LEGENDS legends Color reference

THEMES All the themes you can apply

FACETS facet_grid facet_wrap